capture-ends-with-triagelisted

# A capture pipeline must end with a triage, not a data dump A daily job that captures Teams threads, CI runs, PRs, tickets, and calendar into files is only half a tool. The data is there, but nobody reads ten folders every morning. The operator still has to ask "what needs me today?" by hand, which means the answer arrives late or never. A review request sits unanswered, a red build looks like noise, an @mention scrolls away. The capture is not the deliverable. The **action digest** is. ## The rule Every periodic capture pipeline ends with a triage step that runs last, after capture, and produces a short operator digest: what changed, what is broken, what is waiting on the operator. Persist it (`output/triage/<date>.md`) and print it. ## Triage, not a second pile of data The trap is a digest that just relists the raw data ("here are 16 pipelines, 40 PRs, 200 messages"). That is a second thing to triage. A real triage **resolves each item to a cause and a next action**: - A red CI pipeline is diagnosed by reading its timeline: a failed *task* (needs a code fix, point at the stage) versus an expired/rejected *approval gate* (just re-run and approve, no code change). These look identical in a status list and demand opposite responses. - A PR is surfaced only if the operator is a reviewer who has not voted yet (an actual pending review), not every open PR. - An @mention is the line of text, with its source, filtered to "today", so it reads in seconds. - A work item is sur